Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A method for compressing depth buffer data in a 3-dimensional computer graphics system, wherein the depth buffer data is divided into a plurality of rectangular tiles corresponding to rectangular areas of an associated image, comprising: for each tile to be compressed, identifying a plurality of starting point pixels in the tile; for each starting point pixel of the plurality of starting point pixels: determining a difference in depth between a depth value at the starting point pixel and respective depth values at each of at least two further pixels to determine changes in depth for x and y directions of the tile, and predicting a depth value at a plurality of other pixels in the tile from the determined changes in depth for the x and y directions of the tile, and where a predicted depth value for a pixel substantially matches an actual depth value for that pixel, assigning that pixel to a geometrical plane associated with the starting point pixel; identifying pixels within a tile that are not assigned to any geometrical plane associated with the plurality of starting point pixels; for each tile, storing: for each geometrical plane, starting point pixel and depth value data at the starting point pixel, depth change data, and plane assignment data, the plane assignment data indicating which pixels in the tile are assigned to that geometrical plane, and data indicating which pixels in the tile are not assigned to any of the geometrical planes and depth values for those pixels.
3D computer graphics depth buffer compression. The invention addresses the need to efficiently store depth information in 3D graphics. The method involves dividing depth buffer data into rectangular tiles. For each tile, multiple starting point pixels are identified. From each starting point pixel, changes in depth are calculated in the x and y directions by comparing its depth value to at least two other pixels. These depth changes are then used to predict depth values for other pixels within the tile. If a predicted depth value closely matches the actual depth value of a pixel, that pixel is assigned to a geometrical plane defined by the starting point pixel. Pixels not assigned to any geometrical plane are identified. Finally, for each tile, compressed data is stored. This includes data for each geometrical plane: the starting point pixel and its depth value, the calculated depth change data, and plane assignment data indicating which pixels belong to that plane. Additionally, data for unassigned pixels is stored, including their depth values.
2. A method according to claim 1 in which differences between predicted depth values and actual depth values which are below a threshold are considered to be substantial matches.
This invention relates to depth sensing systems, specifically improving the accuracy of depth prediction by evaluating differences between predicted and actual depth values. The core problem addressed is the challenge of determining when predicted depth values closely match actual depth values, particularly in scenarios where minor discrepancies should not be treated as mismatches. The method involves comparing predicted depth values with actual depth values and identifying substantial matches based on a predefined threshold. If the difference between a predicted depth value and an actual depth value falls below this threshold, the values are considered a substantial match. This approach helps filter out minor variations that may arise due to noise or measurement inaccuracies, ensuring more reliable depth sensing. The threshold can be adjusted based on application requirements, such as the desired precision level or environmental conditions. This method is particularly useful in applications like robotics, augmented reality, and autonomous navigation, where accurate depth perception is critical. By distinguishing between meaningful differences and negligible variations, the system improves the robustness and reliability of depth-based decision-making processes.
3. A method according to claim 2 in which the differences between depth values which are below a threshold are stored as a sparse matrix.
This invention relates to depth sensing systems, specifically methods for efficiently storing and processing depth data. The problem addressed is the computational and memory overhead associated with storing high-resolution depth maps, which often contain redundant or similar depth values. The solution involves a method for compressing depth data by identifying and storing only significant differences between depth values, reducing storage requirements while preserving essential spatial information. The method involves capturing depth data from a depth sensor, such as a time-of-flight camera or structured light system, and analyzing the depth values to identify regions where differences between adjacent or nearby depth values fall below a predefined threshold. These small differences are stored in a sparse matrix format, where only non-zero or significant values are retained, while negligible differences are omitted. This approach minimizes storage space and computational overhead during processing, particularly in applications like 3D reconstruction, object tracking, or augmented reality, where real-time performance is critical. The sparse matrix representation allows for efficient reconstruction of the depth map when needed, as the stored differences can be used to interpolate or reconstruct the full depth information. The threshold for determining significant differences can be dynamically adjusted based on application requirements, ensuring a balance between compression efficiency and data accuracy. This method is particularly useful in scenarios where depth data is processed in real-time or transmitted over bandwidth-limited channels.
4. A method according to claim 1 wherein each of at least two further pixels for each starting point are determined by selecting pixels horizontally and vertically adjacent to that starting point, and comparing depth values at these pixels with the depth value at that starting point pixel.
This invention relates to image processing, specifically depth-based image segmentation or analysis. The problem addressed is efficiently identifying and grouping pixels in an image based on depth values to improve tasks like object detection, scene reconstruction, or 3D modeling. The method involves analyzing depth data from an image to determine connected regions of pixels with similar depth values. For each starting pixel, the method identifies at least two additional pixels by selecting horizontally and vertically adjacent pixels. The depth values of these adjacent pixels are compared to the depth value of the starting pixel. If the depth values are similar, the adjacent pixels are grouped with the starting pixel. This process repeats iteratively to expand the connected region. The method ensures that only pixels with comparable depth values are grouped, improving the accuracy of depth-based segmentation. The technique is useful in applications requiring precise depth-based segmentation, such as autonomous navigation, augmented reality, or medical imaging, where distinguishing objects or surfaces at different depths is critical. By focusing on horizontal and vertical adjacency, the method reduces computational complexity while maintaining segmentation accuracy.
5. A method according to claim 1 wherein the identifying comprises identifying four starting points, located respectively at different corners of each tile.
Technical Summary: This invention relates to image processing, specifically to methods for identifying and analyzing patterns or features within tiled images. The problem addressed is the accurate and efficient detection of key reference points within segmented image regions (tiles) to facilitate further processing, such as alignment, stitching, or feature extraction. The method involves identifying four distinct starting points, each located at different corners of a tile. These points serve as anchor references for subsequent operations. The approach ensures that each tile is analyzed with consistent spatial context, enabling precise alignment or comparison with adjacent tiles. This is particularly useful in applications like image stitching, where misalignment between tiles can lead to artifacts or errors. By defining these corner points, the method provides a structured framework for processing individual tiles while maintaining spatial relationships across the entire image. The technique may involve edge detection, corner detection algorithms, or other image analysis methods to pinpoint these starting points. Once identified, these points can be used to normalize the tile's orientation, scale, or position relative to other tiles. This ensures that subsequent processing steps, such as feature matching or image registration, are performed with high accuracy. The method is applicable in various fields, including medical imaging, satellite imagery, and automated inspection systems, where precise tile alignment is critical.
6. A method according to claim 1 further comprising selecting further starting point pixels in areas, within a particular tile, in which pixels are not assigned to any of the geometrical planes, and the method of claim 1 is repeated so as to assign previously unassigned pixels to further geometrical planes defined based on the further starting point pixels and depth values associated with those further starting point pixels.
This invention relates to image processing, specifically methods for assigning pixels in an image to geometrical planes based on depth information. The problem addressed is the incomplete assignment of pixels to planes, leaving some pixels unassigned in certain areas of an image, particularly within segmented tiles. The solution involves iteratively selecting additional starting point pixels in regions where pixels remain unassigned and repeating the plane-fitting process to incorporate these pixels into new geometrical planes. The method begins by identifying unassigned pixels within a tile of the image. New starting point pixels are then selected from these unassigned regions. Using depth values associated with these starting point pixels, additional geometrical planes are defined. The process is repeated iteratively, ensuring that previously unassigned pixels are progressively incorporated into the plane assignments. This approach improves the completeness of the plane-fitting process, reducing gaps in the reconstructed 3D structure of the image. The technique is particularly useful in applications requiring detailed depth mapping, such as 3D reconstruction, augmented reality, and computer vision tasks where accurate plane assignments are critical.
7. A method according to claim 1 in which the association of pixels to geometrical planes is performed, for multiple of the plurality of starting point pixels in a tile, in parallel.
This invention relates to image processing, specifically methods for associating pixels to geometrical planes in a 3D reconstruction or depth estimation process. The problem addressed is the computational inefficiency of sequentially processing pixels when determining their association with geometric planes, which can slow down real-time applications like augmented reality, autonomous navigation, or 3D modeling. The method involves selecting a plurality of starting point pixels within a tile of an image. For multiple of these starting point pixels, the association of pixels to geometrical planes is performed in parallel. This parallel processing reduces the time required to analyze the image by distributing the computational load across multiple processing units. The geometrical planes may represent surfaces or objects in a 3D space, and the association process involves determining which pixels belong to which planes based on depth or structural information. The method may also include preprocessing steps to identify starting points or regions of interest within the tile, ensuring efficient parallelization. By processing multiple starting points simultaneously, the method improves the speed and scalability of 3D reconstruction tasks, making it suitable for high-resolution images or real-time applications.
8. A method according to claim 1 , in which the association of pixels to geometrical planes is performed, for more than one pixel in a tile, in parallel.
This invention relates to image processing, specifically methods for associating pixels to geometrical planes in a 3D reconstruction or depth mapping system. The problem addressed is the computational inefficiency of sequentially processing pixels to determine their association with geometric planes, which can slow down real-time applications such as augmented reality, robotics, or 3D scanning. The method involves dividing an image into tiles, where each tile contains multiple pixels. For each tile, the system processes more than one pixel simultaneously to determine their association with geometric planes. This parallel processing approach reduces the time required to analyze the entire image by leveraging parallel computing resources, such as multi-core processors or graphics processing units (GPUs). The geometric planes may represent surfaces, objects, or other structural elements in a 3D space, and the associations are used to generate a depth map or 3D model. The method may also include preprocessing steps to optimize the parallel processing, such as identifying regions of interest or simplifying the geometric planes before pixel association. The parallel processing can be implemented using hardware acceleration or specialized algorithms designed for parallel execution. This approach improves processing speed without sacrificing accuracy, making it suitable for applications requiring real-time 3D reconstruction.
9. A method according to claim 1 further comprising compressing and storing stencil values along with the depth values.
The invention relates to a method for processing and storing stencil values in a graphics rendering system. The method addresses the challenge of efficiently managing stencil data alongside depth values during rendering operations, which is critical for improving performance and memory usage in real-time graphics applications. The method involves compressing stencil values and storing them in conjunction with depth values. This approach optimizes storage efficiency by reducing the memory footprint of stencil data, which is particularly beneficial in systems where memory bandwidth and storage capacity are limited. By compressing the stencil values, the method ensures that the additional overhead of storing stencil information does not significantly impact rendering performance. The compression technique used may involve lossless or lossy methods, depending on the specific requirements of the application. The compressed stencil values are stored in a manner that allows for efficient retrieval and decompression during rendering, ensuring that the stencil test operations can be performed without noticeable delays. This method is particularly useful in scenarios where multiple rendering passes or complex stencil operations are required, such as in shadow rendering, depth peeling, or other advanced rendering techniques. The invention builds upon a base method for processing depth values, integrating stencil value compression to enhance overall system efficiency. The combined storage of compressed stencil and depth values ensures that both types of data are managed cohesively, reducing memory access latency and improving rendering throughput. This approach is applicable to various graphics processing units (GPUs) and rendering pipelines, making it a versatile solution for m
10. A method according to claim 9 in which the compression of stencil values is performed using run length encoding.
A method for compressing stencil values in a graphics processing system involves encoding the values using run length encoding (RLE) to reduce memory usage and improve processing efficiency. Stencil values are commonly used in computer graphics to control pixel rendering, but storing them in uncompressed form consumes significant memory, especially in high-resolution applications. The method addresses this by identifying sequences of identical stencil values and encoding them as a single value followed by a count, significantly reducing the data size. The compression process is applied to stencil data stored in a buffer, where the buffer may be part of a frame buffer or a dedicated stencil buffer. The method ensures that the compressed data can be efficiently decompressed during rendering operations, maintaining performance while minimizing memory overhead. This approach is particularly useful in real-time rendering applications where memory bandwidth and storage efficiency are critical. The compression technique is applied to stencil values that are part of a larger rendering pipeline, ensuring compatibility with existing graphics processing workflows. The method may also include additional optimizations, such as adaptive compression based on the stencil data characteristics, to further enhance efficiency.
11. A method for decompressing depth buffer data compressed using the method of claim 1 comprising, for each tile: reading the compressed depth data for all geometrical planes in the tile; using plane assignment data from the compressed depth data to determine which pixels in the tile are assigned to any of the geometrical planes; reading any uncompressed depth data for pixels in the tile that were not assigned to any of the geometrical planes; extrapolating, for each pixel in the tile that was assigned to a geometrical plane, depth values for those pixels from the compressed depth data, based on a respective depth at a starting point of a geometrical plane to which that pixel belongs and the depth change data; and providing a set of uncompressed depth values for the pixels in the rectangular tile.
The invention relates to a method for decompressing depth buffer data that has been compressed using a specific compression technique. Depth buffers are used in computer graphics to store depth information for each pixel in a scene, and compressing this data is essential for efficient storage and transmission. The problem addressed is the need for an efficient decompression method that accurately reconstructs the original depth values from compressed data. The compression method involves dividing the depth buffer into tiles and representing depth values within each tile using geometrical planes. Each plane is defined by a starting depth value and depth change data, which describes how the depth varies across the plane. Pixels not assigned to any plane are stored in uncompressed form. The decompression method processes each tile individually. First, it reads the compressed depth data for all geometrical planes in the tile. Using plane assignment data, it determines which pixels in the tile are assigned to any of the geometrical planes. For pixels not assigned to a plane, the method reads the uncompressed depth data. For pixels assigned to a plane, it extrapolates depth values based on the starting depth value and the depth change data of the respective plane. Finally, the method provides a set of uncompressed depth values for all pixels in the tile, reconstructing the original depth buffer. This approach ensures efficient decompression while maintaining accuracy in the reconstructed depth values.
12. A method according to claim 1 , wherein, for a tile, at least two different pixels are respectively assigned to at least two different geometrical planes.
This invention relates to image processing, specifically techniques for rendering or displaying images with enhanced depth perception. The problem addressed is the lack of realism in flat, two-dimensional images, which fail to convey the spatial relationships and depth cues present in real-world scenes. The solution involves assigning different pixels within a single image tile to different geometrical planes, creating a multi-plane representation that simulates depth. The method processes an image by dividing it into tiles, where each tile contains multiple pixels. For at least one tile, at least two distinct pixels are assigned to different geometrical planes. This means that within a single tile, some pixels are rendered at a different depth or distance from the viewer than others, creating a layered effect. The assignment of pixels to different planes can be based on depth information extracted from the image, such as disparity maps, depth maps, or other depth estimation techniques. The result is an image that appears more three-dimensional, with objects or regions at varying depths, improving visual realism and user experience. This technique can be applied in various applications, including virtual reality, augmented reality, 3D displays, and advanced image rendering systems. By introducing depth variation within individual tiles, the method enhances the perception of depth without requiring complex hardware or excessive computational resources. The approach is particularly useful for improving the quality of images in environments where depth cues are critical, such as medical imaging, gaming, and immersive media.
13. Apparatus for compressing depth buffer data in a 3-dimensional computer graphics system comprising: a depth buffer compressor configured to identify a plurality of starting point pixels for each tile, of a plurality of tiles in an image, for which depth data is to be compressed; a z buffer compressor configured to: for each starting point pixel of the plurality of starting point pixels in each tile, determine a difference in depth between a depth value at the starting point pixel of the plurality of starting point pixels in the tile, and depth values at a respective plurality of other pixels in that tile to determine changes in depth for x and y directions of the tile; predict a depth value at a plurality of other pixels in the tile from the determined changes in depth for the x and y directions of the tile; determine whether a predicted depth value substantially matches an actual depth value for each pixel of said plurality of other pixels, and where there is a substantial match assigning that pixel to a geometrical plane associated with the starting point pixel; and identify pixels in a tile that are not assigned to any of the geometrical planes; and store, for each tile: for each of the geometrical planes, the starting point pixel and depth data, depth change data, and plane assignment data, the plane assignment data indicating which pixels in the tile are assigned to that geometrical plane, and depth values for pixels in the tile not assigned to any geometrical plane.
This apparatus compresses depth buffer data in 3D computer graphics systems to reduce memory usage and bandwidth. The system addresses the challenge of efficiently storing depth information for large 3D scenes, where uncompressed depth buffers consume significant memory and processing resources. The apparatus includes a depth buffer compressor and a z buffer compressor. The depth buffer compressor identifies starting point pixels for each tile in an image, where depth data will be compressed. The z buffer compressor processes each starting point pixel by calculating depth differences between the starting pixel and other pixels in the tile, determining depth changes in both x and y directions. It then predicts depth values for other pixels based on these changes and compares predicted values to actual depth values. If a predicted value closely matches the actual value, the pixel is assigned to a geometrical plane associated with the starting point. Pixels that do not match any plane are identified separately. The system stores compressed data for each tile, including starting point pixels, depth values, depth change data, and plane assignment data for pixels assigned to geometrical planes. Unassigned pixels are stored with their actual depth values. This approach reduces storage requirements by leveraging spatial coherence in depth data, encoding only necessary information for reconstruction.
14. Apparatus according to claim 13 in which differences between predicted depth values and actual depth values which are below a threshold are considered to be substantial matches.
This invention relates to depth sensing systems, particularly for comparing predicted depth values with actual depth values to determine matches. The system addresses the challenge of accurately assessing depth data in applications such as robotics, augmented reality, or 3D imaging, where precise depth measurements are critical. The apparatus includes a depth sensor that captures actual depth values from a scene and a processing unit that generates predicted depth values based on a model or prior data. The system compares these values to identify substantial matches, where differences below a predefined threshold are considered acceptable. This threshold-based approach improves robustness by filtering out minor discrepancies that may arise from noise or sensor limitations. The apparatus may also include calibration mechanisms to adjust the threshold dynamically based on environmental conditions or sensor performance. By focusing on substantial matches, the system enhances reliability in depth-based applications, ensuring accurate spatial awareness and interaction. The invention is particularly useful in scenarios requiring real-time depth analysis, such as autonomous navigation or object recognition, where minor variations in depth readings should not trigger false negatives or positives.
15. Apparatus according to claim 14 in which the differences between depth values which are below a threshold are stored as a sparse matrix.
This invention relates to a system for processing depth data, particularly for efficiently storing and representing depth information in a sparse format. The system addresses the challenge of handling large volumes of depth data, which can be computationally intensive and memory-consuming when stored in a dense format. By identifying and storing only significant differences in depth values, the system reduces storage requirements and processing overhead. The apparatus includes a depth sensor that captures depth data representing distances from the sensor to objects in a scene. A processing unit analyzes the depth values to identify regions where the differences between adjacent depth values fall below a predefined threshold. These small differences are stored in a sparse matrix format, where only non-zero or significant values are retained, while negligible differences are omitted. This approach minimizes data redundancy and improves efficiency in subsequent processing tasks such as object recognition, scene reconstruction, or depth-based navigation. The sparse matrix representation allows for faster computation and lower memory usage, making the system suitable for real-time applications in robotics, augmented reality, and autonomous vehicles. The threshold for determining significant depth differences can be dynamically adjusted based on application requirements or environmental conditions. The system may also include additional components for preprocessing raw depth data, such as noise filtering or interpolation, to enhance accuracy before sparse storage. The overall design optimizes depth data handling while maintaining the necessary precision for accurate spatial analysis.
16. Apparatus according to claim 13 in which the depth buffer compressor is further configured for selecting the respective plurality of other pixels as pixels that are horizontally and vertically adjacent to the starting point, and to compare depth values at these pixels with the depth value at the starting point pixel.
This invention relates to depth buffer compression in graphics processing, specifically addressing the challenge of efficiently compressing depth data to reduce memory bandwidth and storage requirements while maintaining accuracy. The apparatus includes a depth buffer compressor that processes depth values stored in a depth buffer, which represents the distance of surfaces from a viewpoint in a 3D scene. The compressor identifies a starting point pixel in the depth buffer and selects a plurality of other pixels that are horizontally and vertically adjacent to this starting point. The depth values at these adjacent pixels are then compared with the depth value at the starting point pixel. This comparison helps determine spatial coherence in the depth data, allowing for more efficient compression by leveraging similarities between neighboring pixels. The compressor may further encode the depth values based on these comparisons, reducing redundancy and improving compression efficiency. The invention aims to optimize memory usage and processing speed in real-time rendering applications by minimizing the data required to represent depth information.
17. Apparatus according to claim 13 in which the plurality of starting point pixels are at corners of each tile.
This invention relates to image processing, specifically a method for analyzing or processing images by dividing them into smaller regions called tiles and identifying key pixels within those tiles. The problem addressed is efficiently determining starting points for further image analysis, such as feature detection or segmentation, by selecting specific pixels within each tile. The apparatus includes a processor configured to divide an input image into a grid of tiles and identify a plurality of starting point pixels within each tile. These starting point pixels are positioned at the corners of each tile, ensuring uniform distribution across the image. The processor then performs further image analysis operations using these starting points, such as feature extraction, edge detection, or object segmentation. The method ensures that the analysis is systematic and covers the entire image without redundancy, improving computational efficiency and accuracy. The apparatus may also include memory for storing the image data and intermediate results, as well as input and output interfaces for receiving and displaying the processed image. The invention is particularly useful in applications requiring real-time image processing, such as computer vision systems, medical imaging, and autonomous navigation.
18. Apparatus according to claim 13 in which the depth buffer compressor is configured to select further starting point pixels in areas of unassigned pixels within a tile.
This invention relates to graphics processing, specifically to techniques for compressing depth buffers in tile-based rendering systems. The problem addressed is the inefficiency in memory usage and bandwidth when storing depth information for large numbers of pixels, particularly in areas with sparse or unassigned pixel data within a tile. The apparatus includes a depth buffer compressor that processes depth data for pixels within a tile, which is a fixed-size block of pixels in a frame buffer. The compressor identifies starting point pixels within the tile, which serve as reference points for encoding depth values of neighboring pixels. To improve compression efficiency, the compressor is configured to select additional starting point pixels in areas where pixels remain unassigned, ensuring that depth data is accurately represented even in regions with sparse coverage. This approach reduces the amount of data that needs to be stored or transmitted while maintaining the integrity of the depth buffer. The system may also include a depth buffer decompressor that reconstructs the original depth values from the compressed data using the selected starting points. The compressor and decompressor work together to minimize memory usage and bandwidth requirements, particularly in scenarios where only a subset of pixels within a tile are active or require depth information. This technique is useful in graphics rendering pipelines where efficient memory management is critical for performance.
19. Apparatus according to claim 13 in which the depth buffer compressor is configured to assign each starting pixel a precedence value.
A system for depth buffer compression in computer graphics processes depth data to reduce memory usage while maintaining accuracy. The system addresses the challenge of efficiently storing and transmitting depth information in real-time rendering applications, where large amounts of depth data can consume significant memory and bandwidth. The apparatus includes a depth buffer compressor that assigns a precedence value to each starting pixel in a depth buffer. This precedence value determines the order in which pixels are processed during compression, ensuring that higher-priority pixels are preserved with greater accuracy. The compressor may also use a hierarchical approach, where pixels are grouped into blocks or regions, and compression is applied based on spatial coherence within those regions. The system further includes a decompression module that reconstructs the depth buffer from compressed data, maintaining sufficient precision for rendering tasks. The apparatus may be integrated into graphics processing units (GPUs) or other hardware accelerators to optimize performance in applications such as virtual reality, gaming, and 3D visualization. The compression technique reduces memory footprint without sacrificing visual quality, enabling smoother rendering and improved system efficiency.
20. Apparatus according to claim 19 in which the assignment of pixels to geometrical planes is performed for more than one of the plurality of starting points in a tile in parallel, and the depth buffer compressor is further configured to reassign a pixel already assigned to a geometrical plane to a different geometrical plane whose starting pixel has a higher precedence.
This invention relates to a system for compressing depth buffers in computer graphics, particularly for improving rendering efficiency in real-time applications. The problem addressed is the computational overhead and memory usage associated with storing and processing depth information for complex 3D scenes, which can degrade performance in graphics rendering pipelines. The apparatus includes a depth buffer compressor that assigns pixels to multiple geometrical planes based on their depth values. Each plane is defined by a starting point within a tile, and pixels are assigned to the nearest plane. The system processes multiple starting points in parallel within a tile to accelerate the assignment process. Additionally, the compressor dynamically reassigns pixels already assigned to one plane to a different plane if the new plane's starting pixel has higher precedence, ensuring optimal compression and accuracy. This reassignment step helps maintain consistency and reduces artifacts in the rendered output. The invention improves efficiency by leveraging parallel processing and dynamic reassignment, reducing memory bandwidth and computational load while preserving visual quality. This is particularly useful in applications requiring high-performance rendering, such as video games, virtual reality, and real-time 3D simulations. The system balances speed and accuracy, making it suitable for hardware-accelerated graphics pipelines.
21. Apparatus according to claim 13 in which the assignment of pixels to geometrical planes is performed for more than one pixel in a tile in parallel.
This invention relates to image processing systems that assign pixels to geometrical planes for depth estimation or 3D reconstruction. The problem addressed is the computational inefficiency of processing pixels sequentially, which limits performance in real-time applications. The apparatus includes a pixel processing unit that assigns pixels to geometrical planes in parallel for multiple pixels within a tile. A tile is a predefined group of pixels, and the system processes these pixels concurrently rather than sequentially. This parallel processing improves throughput by leveraging hardware acceleration, such as parallel processing units or specialized circuits. The system may also include a memory interface to fetch pixel data and a control unit to manage the parallel assignment process. The geometrical planes are predefined or dynamically determined based on scene characteristics. The parallel assignment reduces latency and power consumption compared to sequential methods. The apparatus may further include error correction mechanisms to handle misassignments due to parallel processing. This approach is particularly useful in applications like 3D imaging, autonomous navigation, and augmented reality, where real-time performance is critical. The invention enhances efficiency without sacrificing accuracy, making it suitable for high-resolution or high-frame-rate systems.
Unknown
August 20, 2019
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.